Research On The Evaluation Method For Image Geometric Precision Of Domestic Optical Satellites

In recent years,high resolution optical satellite technology has a great development in our country,has launched ZY-301 ? GF-1 ? ZY-302 and so on.It is the production,installation,orbit,attitude and the influence of the imaging target itself that make the original satellite image produce distortion,so that it is necessary to use the error model to improve the geometric positioning accuracy of remote sensing images.Usually geometric positioning models can be divided into two types,one is the Rigorous Geometrical Model,and the other is the Rational Function Model.The Rigorous Geometrical Model can be used to describe the internal and external distortion of image and the law of image positioning error,at the same time as the parameters of the Rigorous Geometrical Model belongs to the confidential information,so the Rational Function Model instead of rigorous model,the Rational Function Model simplified the Rigorous Geometrical Model,it is merely a description of image in line and row direction.At present,the Rational Function Model and the Rigorous Geometrical Model is progressing rapidly.Using different models to evaluate the geometric accuracy of remote sensing satellite images is particularly important,it also can provide the basis for the geometry precision of image compensation.In this paper,based on the imaging geometry of home made optical satellite linear pushbroom sensor,the analysis of error sources and the rules based on the Rigorous Geometrical Model.The sources of satellite error are divided into attitude error,orbit error and line time error,and simulation experiments analysis are carried out respectively.The attitude error is mainly composed of satellite rolling angle,pitch angle and yaw angle,which directly affect the external accuracy of the image.By using the RFM substitution model to analyze the CCD distortion,backward inference the interior image pixel stitching accuracy of the image,and obtain the influence of the internal image accuracy by the CCD distortion.The system error is obtained by experiments on GF-1 image.By using four different error models,different evaluation methods and evaluation indexes,the geometric accuracy of ZY301,ZY302 and GF-1 images are evaluated.In contrast to the different evaluation indexes,error model and experimental data,the significance of each index and the difference of different model evaluation methods are expounded.The traditional method of geometric accuracy evaluation is easy to calculate.The method selects a part of all points as the control point,and the rest as the checkpoint to evaluate the geometric accuracy of the image.The number of control points and the quality of control points determine the evaluation accuracy,and it is difficult to make reasonable evaluation in the areas where the control points are scarce.Therefore,this paper uses the LOOCV method to evaluate the geometric accuracy of the image.The selected points are both the control point and the checkpoint,which are used to improve the geometric accuracy of the region and evaluate the quantity and quality of the control points.Therefore,this paper uses the leave aside method to evaluate the geometric accuracy of the image.The selected points are both the control point and the checkpoint,which are used to improve the number and quality of the control points used in the geometric accuracy evaluation of the region.Through the contrast experiment between the two polynomial model and the affine transformation model,it is proved that LOOCV method can more reasonably evaluate the geometric accuracy of the optical satellite image.